Tentative fix of the signature implementation

[Baccata]

The sigv4 trait contains what seems to be the service-specific name that should be used during the signature of the request.

[kubukoz]

Think we've got this implemented already? AWS seems to work fine

[Baccata]

I think it's still needed, in the sense that it's probably working fine by accident, because the sigv4 name is coincidentally the same as the endpoint prefix for most services

[Baccata]

Anyhow, we need parity tests with existing SDKs

[kubukoz]

oh I see. yes that makes sense